Tectonophysics 632 (2014) 151–159

Contents lists available at ScienceDirect

Tectonophysics

j ourna l homepage: www.e lsev ie r .com/ locate / tecto

Moho depth variations in the Taiwan orogen from joint inversion ofseismic arrival time and Bouguer gravity data

Zhiwei Li a,⁎, Steven Roecker b, Kwanghee Kim c, Ya Xu d, Tianyao Hao d

a State Key Laboratory of Geodesy and Earth's Dynamics, Institute of Geodesy and Geophysics, Chinese Academy of Sciences, Wuhan 430077, Chinab Department of Earth and Environmental Sciences, Rensselaer Polytechnic Institute, Troy, NY 12180, USAc Department of Geological Sciences, Pusan National University, Busan 609-735, Republic of Koread Key Laboratory of Oil and Gas Resources, Institute of Geology and Geophysics, Chinese Academy of Sciences, Beijing 100029, China

⁎ Corresponding author. Tel.: +86 27 86780717.E-mail address: zwli@whigg.ac.cn (Z. Li).

http://dx.doi.org/10.1016/j.tecto.2014.06.0090040-1951/© 2014 Elsevier B.V. All rights reserved.

a b s t r a c t

a r t i c l e i n f o

Article history:Received 4 March 2014Received in revised form 3 June 2014Accepted 9 June 2014Available online 16 June 2014

Keywords:TomographyJoint inversionSeismic arrival timeBouguer gravity anomalySubductionContinental crust

The joint inversion of different geophysical datasets is an effective means to reduce the non-uniqueness andimprove the reliability of geophysical inversion. In this study, seismic arrival time and Bouguer gravity datasetsare jointly inverted to obtain an image of 3-D velocity structures in the Taiwan orogen. The model obtainedfrom joint inversion fits the arrival time observations at least as well as when inverted individually, and thegravity observations aremuch better fit when included in the inversion, implying a reduction in ambiguity by si-multaneously modeling the disparate datasets. Moho depth variations estimated by the 3-D P wave velocitymodel suggest a maximum Moho depth of 56 km located beneath the Backbone Central Range, and the trendof the Moho is largely consistent with the topography of the Central Range with eastward, asymmetric crustalthickening. The root beneath the Central Range appears to be smaller in lateral extent than previously imaged,and the velocity gradients into the uppermost mantle are significantly higher. The lack of evidence for a signifi-cant amount of Eurasian crust in the mantle supports geodynamical models of accreted, rather than consumed,continental crust in a collisional environment.

© 2014 Elsevier B.V. All rights reserved.

1. Introduction

The Taiwan orogen is the result of oblique collision between theEurasian plate (EAP) and the Philippine Sea plate (PSP) thatcommenced about 5 Ma ago (e.g., Wu et al., 1997) (Fig. 1). The fastconvergence rate (~9 cm/yr) across the plate boundary between theEAP and PSP is responsible for the oceanic and continental platecollision and the significantly thickened crust of the Taiwan orogen(e.g., Sella et al., 2002). The EAP subducts eastward under the PSP inthe south along the northern Luzon arc and the PSP subducts north-westward under the EAP along the Ryukyu trench in the north ofTaiwan. This bidirectional subduction results in one of the most activeand complex orogens in the world. The high rate of seismicity anddense seismological and geodetic networks make Taiwan an excellentplace for investigating ongoing orogenic processes in an arc–continentcollision environment. Studies based on gravity anomalies (Hsiehet al., 2010; Yen et al., 1998), seismic arrival times (Kim et al., 2005;Kuo-chen et al., 2012; Lin et al., 1998; Ma et al., 1996; Rau and Wu,

1995; Roecker et al., 1987; Wang et al., 2006; Wu et al., 1997, 2007),Moho reflection and refraction phases (Hsu et al., 2011; Liang et al.,2007), receiver functions (Kim, et al., 2004; Wang et al., 2010a,b) andactive source seismic sounding (McIntosh et al., 2005, 2013; Wanget al., 2004) have been conducted to investigate the crustal structure be-neath the Taiwan orogeny. A key objective of these studies has been todetermine the depth of theMoho, as variations in crustal thickness pro-vide important boundary conditions on mass balance at a convergentboundary, including the ultimate fate (e.g., consumption vs. accretion)of Eurasian crust as it encounters the Luzon arc. To first order, the thick-ening of the crust beneath the Taiwan orogeny, and the participation ofthe lower crust and upper mantle in the orogenic process, have beenwell established by both seismic arrival time tomography (Kim et al.,2005; Kuo-chen et al., 2012;Ma et al., 1996; Rau andWu, 1995; Roeckeret al., 1987; Wu et al., 1997, 2007) and 2-D and 3-D gravity modeling(Hsieh et al., 2010; Mouthereau and Petit, 2003; Wang et al., 2004;Yen et al., 1998; Zhang et al., 2004). However, the depth of theMohobe-neath Taiwan remains controversial and significant discrepancies existfrom the analyses of different datasets using different techniques.While most studies agree that the crust is thickest beneath the CentralRange, estimates of maximum Moho depth range from 33 km fromgravity data inversion (Yen et al., 1998), 38 km from PmP traveltime in-version (Hsu et al., 2011), less than 54 km from receiver functions

120˚ 121˚ 122˚ 123˚

22˚

23˚

24˚

25˚

Philippine Seaplate

Eur

asia

n pl

ate

Man

ila T

ren

chRyu

kyu Tre

nch

Taiw

an S

trai

t~90 mm/y

Taiw

anL

uzo

n A

rc

Nor

th L

uzo

n A

rc

Ryukyu Trench

Coa

stal

Pla

in

Coa

stal

Ran

ge

Wes

tern

Foo

thill

sW

este

rn F

ooth

ills

Eas

tern

Cen

tral

Ran

ge

Hsu

ehsh

an R

ange

Bac

kbon

e R

ange

Longitudinal Valley

Hualien

Taiw

an S

trai

t

Lishan fault

Hengchun Peninsula

Taitung

Fig. 1. Topographic map and themajor tectonostratigraphic belts of Taiwan orogen. The Central Range is composed of the Backbone Range and the Eastern Central Range. Location ofmapis shown in the inset at the lower right.

152 Z. Li et al. / Tectonophysics 632 (2014) 151–159

(Wang et al., 2010a,b), 55 km from joint earthquake and active sourcetomography (Kuo-Chen et al., 2012), and 50–70 km from traveltimetomography (Kim et al., 2005; Ma et al., 1996; Wu et al., 1997).Additionally, two Moho discontinuities for the EAP and PSP relatedto the subduction polarity reversal were inferred beneath theTaiwan orogen from local earthquake tomography (Ustaszewskiet al., 2012).

Because of differences in sampling and sensitivity, joint inversionof disparate geophysical datasets often can reduce model non-uniqueness when compared to independent inversion (e.g.,Gallardo and Meju, 2011). Previous investigators have shown thatgravity data can provide additional constraints on the subsurfacestructure when combined with seismic data. Examples includejoint inversion of gravity data with local earthquake arrival timedata (Roecker et al., 2004; Roy et al., 2005; Wang et al., 2014),teleseismic arrival time data (O'Donnell et al., 2011; Tiberi et al.,2003), active source seismic data (Tondi et al., 2000; Vermeeschet al., 2009), and surface wave dispersion curves (Maceira andAmmon, 2009). Since Bouguer gravity data are sensitive mostly todensity inhomogeneities of the crust and variations in the depth ofthe Moho (e.g., Grad et al., 2009), a better subsurface model can beachieved by fitting both seismic arrival time and gravity observa-tions. In this study, we jointly invert seismic arrival time and

Bouguer gravity data in the Taiwan orogen, and obtain a 3-D crustalvelocity model that can fit both sets of observations.

2. Method and data

The travel times used in the seismic forward problem are calculatedusing a finite difference eikonal equation solver in an earth-centeredspherical coordinate frame (e.g., Zhang et al., 2012), that has beenimplemented previously in other tomographic studies (e.g., Kuo-chenet al., 2012; Li et al., 2009, 2013; Roecker et al., 2004, 2006). FollowingRoecker et al. (2006), hypocenters are determined via a grid search ap-proach to ensure a globalminimum in variance of arrival time residuals,but are still included explicitly in the normal equations and determinedsimultaneously with velocity perturbations. The damped LSQR algo-rithm (Paige and Saunders, 1982) is used for solving the systemof linearequations. A coarse sphericalmeshwith approximate spacing of 6 km inhorizontal and 4–20 km in vertical directions is used for specifying themodel and accumulating partial derivatives, and an interpolated finemesh of ~2 km grid spacing is used for the finite difference traveltimecalculation. Bouguer gravity observations are included in the inversionby presuming a functional relationship between P wave velocity anddensity. In this study we adopt the velocity–density relationshipsdetermined by Gardner et al. (1974) and Christensen and Mooney

153Z. Li et al. / Tectonophysics 632 (2014) 151–159

(1995) for sedimentary, basement and mid-lower crustal rocks of con-tinental lithologies. Specifically, these are ρ = 9.89.3 + 289.1 ∗ VP forVP N 6.0 km/s, and ρ = 0.23 ∗ VP

0.25 for VP b 5.5 km/s. A polynomial isused to obtain density in a smooth fashion between 5.5 and 6.0 km/s.The forward calculations of gravity and its partial derivatives are madedirectly in spherical coordinateswith an adoptive numerical integrationof the Gauss–Legendre quadrature (Asgharzadeh et al., 2007; Li et al.,2011). The strategy for simultaneous inversion of seismic arrival timeand gravity data follows that of Roecker et al. (2004). All observationsare included in each iteration, and a scalar factor is used to weightthe relative sensitivities of arrival time and gravity data. An aposteriori moving window average smoothing of perturbationswith 3 × 3 × 3 grids is applied prior to model updates to mitigatepossible artifacts appearing only at 1–2 grids. Finally, we assess thereliability and resolution of the results by conducting syntheticcheckerboard tests.

The seismic arrival time data used in this study were recorded by 98permanent and temporary stations between 1991 and 2010.Most of thearrival time picks are taken from International Seismological Center(ISC), with some of them prior to 1998 provided by Prof. Futian Liu ofthe CAS Institute of Geology and Geophysics. As our focus is the crustalstructure beneath Taiwan, only earthquakes near land are selected

120˚ 121˚

22˚

23˚

24˚

25˚

Fig. 2. Distributions of seismic stations (red tr

(Fig. 2). To improve the reliability of the velocity model we use arrivaltimedata only fromevents that have been recorded at least by 6 stationsand for which traveltime residuals are within ±3.0 s (approximately 3times the standard deviation of the residuals). After selection, 568215 Pand 443706 S arrivals from 45003 earthquakes are used in the tomo-graphic inversion. Good ray path coverage is achieved in the crust anduppermost Moho beneath most of Taiwan. Bouguer gravity data usedin the joint inversion are taken from 603 readings with ~10 km spacingor less (Yen et al., 1995), which cover most areas of the Taiwan orogen(Fig. 3).

3. Results

We conduct two independent inversions with the same parameters(i.e., model parameterization, starting model, damping, smoothing) inorder to compare the images derived from arrival time only inversionswith those from joint arrival time/gravity inversions. The starting VP

and VS models for both cases (Fig. 4) are based on one-dimensionalmodels used by Kim et al. (2005) and Wu et al. (2007), slightlyadjusted to reduce arrival time residuals. In both inversions, no sig-nificant residual variance reduction was achieved after 7 iterations.For the arrival time only inversion, the root-mean-square (RMS) of

122˚ 123˚

iangles) and earthquakes (closed circles).

-60−90

Bouguer Gravity (mGal)

120˚ 121˚ 122˚

22˚

23˚

24˚

25˚

Bouguer Gravity

9060300-30

Fig. 3. Bouguer gravity anomaly data used in the joint inversion (Yen et al., 1995).

0

20

40

60

80

100

Dep

th (

km)

1 2 3 4 5 6 7 8 9

Velocity (km/s)

Vs Vp

Fig. 4. The one-dimensional starting models for VP and VS used in both the joint arrivaltime/gravity inversion and the arrival time only inversion.

154 Z. Li et al. / Tectonophysics 632 (2014) 151–159

arrival time residuals decreased from 0.73 s to 0.43 s for P arrivalsand from 0.89 s to 0.50 s for S arrivals. For the joint arrival time/gravity inversion, the RMS of arrival time residuals decreased from0.73 s to 0.42 s for P arrivals and from 0.89 s to 0.50 s for S arrivals.The similarity in misfit shows that both models explain the seismicarrival time data equally well. However, while the RMS of the gravityresiduals increased from 31.6 mGal to 34.8 mGal for the arrival timeonly model, it decreased from 31.6 mGal to 13.0 mGal for the arrivaltime/gravity model.

We evaluate the resolution of the images using checkerboard testswith grid spacing of 6 km and approximate anomaly sizes of 12 km,18 km, 24 km and 30 km (Fig. 5). The patterns and amplitudes of check-erboard models are well recovered under most of Taiwan at depths lessthan about 50 km depth, with the best recovery in themiddle to north-ern Central Range. At depths greater than about 60 km, the anomaliesnear the Central Range are well recovered, largely because of the raypaths from the many deep earthquakes in this area.

4. Discussion

Velocity models derived from arrival time/gravity data and from ar-rival time data onlyfit the arrival time observations equallywell, but theforward gravity anomalies show significant discrepancies (Fig. 6).Since the P wave velocity is related to the gravity observation throughthe VP-density relations, differences in the P wave velocities are morepronounced than S wave velocities (Fig. 7). The main features of thePwave velocity models from joint inversion and arrival time only in-version are virtually the same in the upper 40 km (Fig. 7), with dif-ferences increasing at depths greater than 50 km. While thepatterns of velocity anomalies are similar below 50 km, the scaleand amplitude of the low P wave velocity anomalies beneath theCentral Range are smaller and the gradient to the high P wavevelocities east of the Taiwan orogen becomes larger in the arrivaltime/gravity model. The primary reason for these differences is thatthese regions are relatively poorly sampled by the arrival time data.Hence these velocities can change without much effect on the arrivaltime misfit.

Comparing the arrival time only and joint arrival time/gravity im-ages (Fig. 7), it is clear that the main advantage of including Bouguergravity data in a joint inversion is that it significantly improves the def-inition of the Moho, particularly at depths greater than about 40 km.Choosing a contouring surface of VP = 7.5 km/s as a means to estimatetheMohodepth beneath the Taiwan orogen (Kuo-Chen et al., 2012), themain difference is that the root under the Central Range is narrower,and the gradients in the upper mantle sharper, than the relativelydeeper and broader root in the arrival timemodel. Note that the arrivaltime only model suggests the presence of low wavespeed materialextending both to the base of the model at 60 km depth and to theeast at depths between 40 and 60 km depth. One might infer fromthis image the subduction of the Eurasian continental crust as a meansto obtain these low wavespeed anomalies. The joint inversion model,however, shows that this extended lowwavespeed zone is largely an ar-tifact that results in themisfit in the gravity observations along the east

120˚00' 120˚30' 121˚00' 121˚30' 122˚00'

22˚00'

22˚30'

23˚00'

23˚30'

24˚00'

24˚30'

25˚00'

Depth :50 km Vp

7.6 7.8 8.0 8.2

~30x30km

Vp (km/s)

120˚00' 120˚30' 121˚00' 121˚30' 122˚00'

22˚00'

22˚30'

23˚00'

23˚30'

24˚00'

24˚30'

25˚00'

Depth :20 km Vp

6.2 6.4 6.6

~24x24km

Vp (km/s)

Fig. 5.Representative results from the checkerboard tests of resolution at 20 kmand50kmdepths. A larger scale anomaly is taken at 50kmbecause of the reduced level of resolution at thisdepth.

155Z. Li et al. / Tectonophysics 632 (2014) 151–159

coast of Taiwan (Fig. 6). This joint inversion model therefore arguesagainst the subduction of significant amounts of Eurasian crust intothe mantle, as has been suggested for continental collisions in centralAsia (e.g., Roecker, 1982; Schneider et al., 2013; Sippl et al., 2013), butsupports geodynamical models for crustal accretion.

The main features of velocity structure that we find in the upper40 km are similar to those reported in previous local earthquake tomog-raphy studies (e.g., Kim et al., 2005; Wu et al., 2007). These include thelow-VP, low-VS and high-VP/VS anomalies found in the Coastal Plain at 0and 4 km depth, which extend eastward to the Western Foothills at 8and 12 km depth, and the low-VP, low-VS and normal-to-low VP/VS be-neath the Central Range that become more predominant at 16 to40 km depth.

Previous studies also have suggested that the crust has thickened tomore than 50 km beneath the Central Range (e.g., Kim et al., 2005; Rauand Wu, 1995; Wu et al., 2007). The asymmetrically thickened crustshown clearly in the near E–W profiles (Fig. 7a–d) is similar to thatobtained from joint local and teleseismic tomographic results (e.g.,Kuo-Chen et al., 2012). The Moho depth from the arrival time/gravitymodel (Fig. 8) is close to the estimated Moho depth obtained by Kuo-Chen et al. (2012) using the denser seismic stations and active sourcesof TAIGER project, although we note that, as with our arrival time onlymodel, their model does not fit the Bouguer gravity data particularlywell (their model misfits the east–west trend in gravity by a total ofabout 100 mGal). Comparison with the crustal thickness results fromH–κ stacks of receiver functions (Wang et al., 2010a,b) gives Mohodepths within a few kilometers for most stations in the central part ofthe Taiwan orogeny (Fig. 8). Note that the large differences are locatedmostly in the eastern and northern margin of Taiwan, where the crustand uppermantle appear to be complicated by the history of subductionin this area— for example, twoMoho discontinuities of the EAP and PSPwere proposed to exist here as a result of the change in subduction

polarity (Ustaszewski et al., 2012). Such complications could make theidentification of interfaces producing mode conversions problematic.Finally, the Moho map determined by joint inversion is similar to thatobtained by Zhang et al. (2004), who carried out a crustal structure in-version using gravity datawith a 3-D seismic velocitymodel as an initialmodel, but it is not clear how compatible theirmodel is with the seismicobservations.

5. Conclusions

We have carried out a joint inversion of seismic arrival time andgravity data in the Taiwan orogen. The 3-D velocity model obtainedfrom the joint inversion fits both seismic arrival time and gravityobservations at approximately the noise level of the observations.Comparisons between the models from joint inversion and arrivaltime only inversion show that the well resolved velocity structuresremain almost the same for regions with good ray path coverage.Gravity data provide additional constraints on the velocity structurenear the Moho and in the uppermost mantle where seismic ray pathsare relatively sparse. Hence, the reliability of the Moho depth esti-mated from the velocity model is improved by the joint inversionof seismic arrival time and Bouguer gravity data. The main resultfrom the image generated by this study is that the root under theCentral Range is narrower, and the gradients in the upper mantlesharper than previously determined. This model therefore favorsthe accretion of Eurasian crust, as opposed to subduction of signifi-cant amounts of lower crust into the mantle.

Acknowledgments

We thank the editor, Prof. Laurent Jolivet, and two anonymous re-viewers for their valuable comments. We thank Prof. Sidao Ni for

120˚ 121˚ 122˚

22˚

23˚

24˚

25˚Calculated

120˚ 121˚ 122˚

22˚

23˚

24˚

25˚Residual

−90 −60 −30 0 30 60 90

Bouguer Gravity (mGal)

120˚ 121˚ 122˚

22˚

23˚

24˚

25˚Calculated

120˚ 121˚ 122˚

22˚

23˚

24˚

25˚Residual

(a) (b)

(c) (d)

A

A’

Joint Inversion

Joint Inversion

Arrivaltime only

Arrivaltime only

Fig. 6. (a) Calculated gravity anomaly data from the joint arrival time/gravity model and (b) associated residuals from the Bouguer gravity observations; (c) calculated gravity anomalydata from arrival time only model and (d) associated residuals from the Bouguer gravity observations.

156 Z. Li et al. / Tectonophysics 632 (2014) 151–159

helpful discussions. This work was supported by NSFC41210005,NSFC41304045, XDB06030203 and SKLGED2013-9-3-Z. Participa-tion of S. Roecker in this study was supported by NSF Grant EAR-1010580. H. Kim in this study was funded by the KMA, Research De-velopment Program under Grant CATER 2014-5030.

References

Asgharzadeh, M.F., von Frese, R.R.B., Kim, H.R., Leftwich, T.E., Kim, J.W., 2007. Sphericalprism gravity effects by Gauss–Legendre quadrature. Geophys. J. Int. 169, 1–11.

Christensen, N.I., Mooney, W.D., 1995. Seismic velocity structure and composition of thecontinental crust: a global view. J. Geophys. Res. 100, 9761–9788.

0

10

20

30

40

50

60

70

80

Dep

th (

km)

120.0 120.5 121.0 121.5 122.0Longitude (deg)

6.57

7.5

−303

Ele

v (k

m) 0 50 100 150 200

Distance (km)

5.0 5.5 6.0 6.5 7.0 7.5 8.0Vp (km/s)

0

10

20

30

40

50

60

70

80

Dep

th (

km)

120.0 120.5 121.0 121.5 122.0Longitude (deg)

3.75

44.25

−303

Ele

v (k

m) 0 50 100 150 200

Distance (km)

2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6Vs (km/s)

0

10

20

30

40

50

60

70

80

Dep

th (

km)

120.0 120.5 121.0 121.5 122.0Longitude (deg)

1.7

−303

Ele

v (k

m) 0 50 100 150 200

Distance (km)

1.60 1.65 1.70 1.75 1.80 1.85Vp/Vs

0

10

20

30

40

50

60

70

80

Dep

th (

km)

120.0 120.5 121.0 121.5 122.0Longitude (deg)

6.57

7.5

−303

Ele

v (k

m) 0 50 100 150 200

Distance (km)

5.0 5.5 6.0 6.5 7.0 7.5 8.0Vp (km/s)

0

10

20

30

40

50

60

70

80

Dep

th (

km)

120.0 120.5 121.0 121.5 122.0Longitude (deg)

3.754

4.25

−303

Ele

v (k

m) 0 50 100 150 200

Distance (km)

2.8 3.0 3.2 3.4 3.6 3.8 4.0 4.2 4.4 4.6Vs (km/s)

0

10

20

30

40

50

60

70

80

Dep

th (

km)

120.0 120.5 121.0 121.5 122.0Longitude (deg)

1.7

1.8

−303

Ele

v (k

m) 0 50 100 150 200

Distance (km)

1.60 1.65 1.70 1.75 1.80 1.85Vp/Vs

A A’ A A’

Joint Inversion Arrival time only VpVp

Joint Inversion

Joint Inversion Arrival time only

VsVs

Vp/VsVp/Vs

(a)

(c)

(e)

(b)

(d)

(f)

Arrival time only

Fig. 7. (a) and (b) show the Pwave velocity variations alongAA’ profile (shown in Fig. 6b) for the joint arrival time/gravitymodel and the arrival time onlymodel, respectively; (c) and (d) arethe same as (a) and (b) for S wave velocity; (e) and (f) are the same as (a) and (b) but for VP/VS. Note that structures off of the eastern coast of Taiwan (i.e., distances greater than about150 km) generally are not well constrained by this inversion.

157Z. Li et al. / Tectonophysics 632 (2014) 151–159

Gallardo, L.A., Meju, M.A., 2011. Structure-coupled multiphysics imaging in geophysicalsciences. Rev. Geophys. 49, RG1003.

Gardner, G.H.F., Gardner, L.W., Gregory, A.R., 1974. Formation velocity and density: the di-agnostic basics for stratigraphic traps. Geophysics 39, 770–780.

Grad, M., Tiira, T., ESCWorking Group, 2009. The Moho depth map of the European plate.Geophys. J. Int. 176, 279–292.

Hsieh, H.H., Yen, H.Y., Shih, M.H., 2010. Moho depth derived from gravity data in theTaiwan Strait area. Terr. Atmos. Ocean. Sci. 21 (2), 235–241.

Hsu, H.J., Wen, S., Chen, C.H., 2011. 3D topography of the Moho discontinuity in theTaiwan area as extracted from travel time inversion of PmP phases. J. Asian EarthSci. 41, 335–343.

Kim, K.H., Chiu, J.M., Kao, H., Liu, Q.Y., Yeh, Y.H., 2004. A preliminary study of crustalstructure in Taiwan region using receiver function analysis. Geophys. J. Int. 159,146–164.

Kim, K.H., Chiu, J.M., Pujol, K., Chen, C., Huang, B.S., Yeh, Y.H., Shen, P., 2005. Three-dimensional VP and VS structural model associated with the active subduction andcollision tectonics in the Taiwan region. Geophys. J. Int. 162, 204–220.

Kuo-Chen, H., Wu, F.T., Roecker, S.W., 2012. Three-dimensional P velocity structure of thelithosphere beneath Taiwan from the analysis of TAIGER and related seismic datasets. J. Geophys. Res. 117, B06306.

Li, Z.W., Roecker, S.W., Li, Z.H., Wei, B., Wang, H.T., Gennady, S., Vitaly, B., 2009. Tomo-graphic image of the crust and upper mantle beneath the western Tien Shan fromthe MANAS broadband deployment: possible evidence for lithospheric delamination.Tectonophysics 477, 49–57.

Li, Z.W., Hao, T.Y., Xu, Y., Xu, Y., 2011. An efficient and adaptive approach for modelinggravity effects in spherical coordinates. J. Appl. Geophys. 73, 221–231.

Li, Z.W., Tian, B.F., Liu, S., Yang, J.S., 2013. Asperity of the 2013 Lushan earthquake in theeasternmargin of Tibetan plateau from seismic tomography and aftershock relocation.Geophys. J. Int. 195, 2016–2022.

Liang, W.T., Chiu, J.M., Kim, K., 2007. Anomalous Pn wave observed in Eastern Taiwan: im-plications of a thin crust and elevated oceanic upper mantle beneath the activecollision-zone suture. Bull. Seismol. Soc. Am. 97, 1370–1377.

Lin, C.H., Huang, B.S., Roecker, S.W., Chiu, J.M., Yeh, Y.H., Yen, H.Y., Chen, K.C., 1998. Three-dimensional elastic wave velocity structure of the Hualien region of Taiwan: evidenceof active crustal exhumation. Tectonics 17 (1), 89–103.

Ma, K.F., Wang, J.H., Zhao, D.P., 1996. Three-dimensional seismic velocity structure of thecrust and uppermost mantle beneath Taiwan. J. Phys. Earth 44, 85–105.

Maceira, M., Ammon, C.J., 2009. Joint inversion of surface wave velocity and gravity obser-vations and its application to central Asian basins shear velocity structure. J. Geophys.Res. 114, B02314.

30

40

40

50

120˚ 121˚ 122˚

22˚

23˚

24˚

25˚

(a)

40

40

40

50

120˚ 121˚ 122˚

22˚

23˚

24˚

25˚

(b)

25 30 35 40 45 50 55 60

Depth (km)

Arrival time onlyJoint inversion

1.2

−7.1

−14.7

−0.2

3.6±1.2−4.2

3.6

4.7

−17.5

−2±1.7

−6.5

−11.6

−1.6±1.8

−5.8±6.0

−10.1±2.5

−14.1±3.1

−3.6±2.0

−14.9±1.6

−4.3±3.41.6

−4.6

−15.7

−0.3

2.1±1.2−4.3

4

3.4

−17.4

−0.9±1.7

−5.5

−11.2

−2.2±1.8

−5.5±6.0

−9.1±2.5

−18±3.1

−5.2±2.0

−15.4±1.6

−10±3.4

Vp=7.50 km/s Vp=7.50 km/s

Fig. 8.Contour surface ofVP=7.5 km/s from the (a) joint arrival time/gravitymodel and (b) arrival time onlymodel, taken as a proxy for the depth toMoho. Colored circles summarize theMoho depth derived from H–κ staking of receiver functions (Wang et al., 2010a,b). White numbers indicate the difference in km between the receiver function and joint inversion esti-mates, alongwith the associated uncertainties (where available) in the receiver function estimates. Note that the area near N23.75°, E120.5° ismasked by gray in the joint inversion resultbecause the difference in models results in this area being sampled by less than 6 rays.

158 Z. Li et al. / Tectonophysics 632 (2014) 151–159

McIntosh, K., Nakamura, Y., Wang, T.K., Shih, R.C., Chen, A., Liu, C.S., 2005. Crustal-scaleseismic profiles across Taiwan and the western Philippine Sea. Tectonophysics 401,23–54.

McIntosh, K.D., Van Avendonk, H.J.A., Lavier, L.L., Lester, R., Eakin, D., Wu, F., Liu, C.-S., Lee,C.-S., 2013. Inversion of a hyper-extended rifted margin in the southern CentralRange of Taiwan. Geology 41, 871–874.

Mouthereau, F., Petit, C., 2003. Rheology and strength of the Eurasian continentallithosphere in the foreland of the Taiwan collision belt: constraints from seismicityflexure, and structure styles. J. Geophys. Res. 108 (B11), 2512.

O'Donnell, J.P., Daly, E., Tiberi, C., Bastow, I.D., O'Reilly, B.M., Readman, P.W., Hauser, F.,2011. Lithosphere–asthenosphere interaction beneath Ireland from joint inversionof teleseismic P-wave delay times and GRACE gravity. Geophys. J. Int. 184,1379–1396.

Paige, C.C., Saunders, M.A., 1982. LSQR: an algorithm for sparse linear equations andsparse least squares. TOMS 8 (1), 43–71.

Rau, R.J., Wu, F.T., 1995. Tomographic imaging of lithospheric structures under Taiwan.Earth Planet. Sci. Lett. 133, 517–532.

Roecker, S.W., 1982. Velocity structure of the Pamir–Hindu Kush region: possibleevidence of subducted crust. J. Geophys. Res. 87 (B2), 945–959.

Roecker, S.W., Yeh, Y.H., Tsai, Y.B., 1987. Three-dimensional P and S wave velocity struc-tures beneath Taiwan-deep structure beneath an arc–continent collision. J. Geophys.Res. 92, 10547–10570.

Roecker, S.W., Thurber, C., McPhee, D., 2004. Joint inversion of gravity and arrival timedata from Parkfield New constraints on structure and hypocenter locations near theSAFOD drill site. Geophys. Res. Lett. 31, L12S04.

Roecker, S.W., Thurber, C., Roberts, K., Powell, L., 2006. Refining the image of the San AndreasFault near Parkfield, California using a finite difference travel time computationtechnique. Tectonophysics 426, 189–205.

Roy, L., Sen, M.K., McIntosh, K., Stoffa, P.L., Nakamura, Y., 2005. Joint inversion of firstarrival seismic travel-time and gravity data. J. Geophys. Eng. 2, 277–289.

Schneider, F., Yuan, X., Schurr, B., Mechie, J., Sippl, C., Haberland, C., Minaev, V.,Oimahmadov, I., Gadoev, M., Radjabov, N., Abdybachaev, U., Orunbaev, S.,Negmatullaev, S., 2013. Seismic imaging of subducting continental lower crustbeneath the Pamir. Earth Planet. Sci. Lett. 375, 101–112.

Sella, G.F., Dixon, T.H., Mao, A., 2002. REVEL: a model for recent plate velocities from spacegeodesy. J. Geophys. Res. 107 (B4), 2081.

Sippl, C., Schurr, B., Yuan, X., Mechie, J., Schneider, F., Gadoev, M., Orunbaev, S.,Oimahmadov, I., Haberland, C., Abdybachaev, U., Minaev, V., Negmatullaev, S.,Radjabov, N., 2013. Geometry of the Pamir–Hindu Kush intermediate-depthearthquake zone from local seismic data. J. Geophys. Res. Solid Earth 118,1438–1457.

Tiberi, C., Diament, M., Deverchere, J., Petit-Mariani, C., Mikhailov, V., Tikhotsky, S.,Achauer, U., 2003. Deep structure of the Baikal rift zone revealed by joint inversionof gravity and seismology. J. Geophys. Res. 108, B32133.

Tondi, R., de Franco, R., Barzaghi, R., 2000. Sequential integrated inversion of refractionand wide-angle reflection traveltimes and gravity data for two-dimensional velocitystructures. Geophys. J. Int. 141, 679–698.

Ustaszewski, K., Wu, Y.M., Suppe, J., Huang, H.H., Chang, C.H., Carena, S., 2012. Crustal–mantle boundaries in the Taiwan–Luzon arc–continent collision system determinedfrom local earthquake tomography and 1D models: implications for the mode ofsubduction polarity reversal. Tectonophysics 578, 31–49.

Vermeesch, P.M., Morgan, J.V., Christeson, G.L., Barton, P.J., Surendra, A., 2009. Three-dimensional joint inversion of traveltime and gravity data across the Chicxulubimpact crater. J. Geophys. Res. 114, B02105.

Wang, T.K., Lin, S.F., Liu, C.S., Wang, C.S., 2004. Crustal structure of the southernmostRyukyu subduction zone: OBS MCS and gravity modeling. Geophys. J. Int. 157,147–163.

Wang, Z., Zhao, D., Wang, J., Kao, H., 2006. Tomographic evidence for the Eurasian litho-sphere subducting beneath south Taiwan. Geophys. Res. Lett. 33, L18306.

Wang, H.L., Zhu, L.P., Chen, H.W., 2010a. Moho depth variation in Taiwan from teleseismicreceiver functions. J. Asian Earth Sci. 37, 286–291.

Wang, H.L., Chen, H.W., Zhu, L.P., 2010b. Constraints on average Taiwan Reference MohoDiscontinuity Model — receiver function analysis using BATS data. Geophys. J. Int.183, 1–19.

Wang, X.S., Fang, J., Hsu, H., 2014. Three-dimensional density structure of the lithospherebeneath the North China Craton and the mechanisms of its destruction.Tectonophysics 610, 150–158.

Wu, F.T., Rau, R.J., Salzberg, D., 1997. Taiwan orogeny: thin-skinned or lithosphericcollision? Tectonophysics 274, 191–220.

Wu, Y.M., Chang, C.H., Zhao, L., Shyu, J.B.H., Chen, Y.G., Sieh, K., Avouac, J.P., 2007. Seismictomography of Taiwan: improved constraints from a dense network of strongmotionstations. J. Geophys. Res. 112, B08312.

159Z. Li et al. / Tectonophysics 632 (2014) 151–159

Yen, H.Y., Yeh, Y.H., Lin, C.H., Chen, K.J., Tsai, Y.B., 1995. Gravity survey of Taiwan. J. Phys.Earth 43, 685–696.

Yen, H.Y., Yeh, Y.H., Wu, F.T., 1998. Two-dimensional crustal structures of Taiwan fromgravity data. Tectonics 17 (1), 104–111.

Zhang, J., Wang, C.Y., Shi, Y.L., Cai, Y.E., Chi, W.C., Dreger, D., Cheng, W.B., Yuan, Y.H.,2004. Three-dimensional crustal structure in central Taiwan from gravity inver-sion with a parallel genetic algorithm. Geophysics 69 (4), 917–924.

Zhang, H., Roecker, S.W., Thurber, C.H., Wang, W., 2012. Seismic imaging ofmicroblocks and weak zones in the crust beneath the southeastern margin ofthe Tibetan Plateau. In: Ahmad Dar, Imran (Ed.), Earth Sciences. ISBN: 978-953-307-861-8.